Insertion loss is the primary signal-quality specification for LAN magnetics. Here's the complete technical reference for measurement, interpretation, and datasheet evaluation.
Definition and Formula
Insertion Loss (IL):
IL = 20 × log10(V_out / V_in) [dB]
Where:
V_in = signal voltage at transformer primary (source side)
V_out = signal voltage at transformer secondary (load side)
Example:
V_in = 100 mV
V_out = 89.1 mV → IL = 20 × log10(0.891) = -1.00 dB
Convention: expressed as magnitude (positive dB, "less than 1.0 dB")
IEEE 802.3 specifies: IL < 1.0 dB for 100BASE-TX
IEEE 802.3 Limits
StandardMax ILFrequency RangeNote10BASE-T1.0 dB100kHz–10MHz100BASE-TX1.0 dB1MHz–100MHz1000BASE-T1.1 dB1MHz–100MHzPer pair; pair balance < 0.5 dB
Frequency Response Shape
python# Conceptual insertion loss vs. frequency (3 regions)
regions = {
"Low-freq rollup": {
"range": "< 1 MHz",
"cause": "OCL too low → magnetizing branch shunts signal",
"fix": "Higher OCL spec (1000µH for 1000BASE-T)"
},
"Passband (flat)": {
"range": "1–100 MHz",
"cause": "N/A — optimal operating region",
"typical_IL": "0.3–0.6 dB for quality parts"
},
"High-freq rollup": {
"range": "> 100 MHz",
"cause": "Leakage inductance + inter-winding capacitance",
"note": "Outside 802.3 spec range but visible on VNA curve"
}
}
Frequency where 3dB rolloff begins:
Low end: f_3dB_low = R_load / (2π × OCL)
At 100Ω load, 1000µH OCL: f_3dB_low = 100 / (2π × 1000e-6) ≈ 15.9 kHz
import math
R = 100 # Ω
OCL_100M = 350e-6 # 350µH for 100BASE-TX
OCL_1G = 1000e-6 # 1000µH for 1000BASE-T
f_low_100M = R / (2 * math.pi * OCL_100M)
f_low_1G = R / (2 * math.pi * OCL_1G)
print(f"100BASE-TX low-end -3dB: {f_low_100M/1e3:.1f} kHz")
print(f"1000BASE-T low-end -3dB: {f_low_1G/1e3:.1f} kHz")
100BASE-TX: ~45.5 kHz
1000BASE-T: ~15.9 kHz (lower rolloff → supports lower signal frequencies)
Physical Loss Mechanisms
Source Frequency Dependence Effect
──────────────────────────────────────────────────────────────────
Winding DCR Flat (rises with temp) IL floor at all freqs
Core hysteresis Increases with freq Mid/HF loss
Eddy currents f² dependence HF loss
Leakage inductance Increases with freq HF rolloff
Inter-winding cap. Increases with freq HF bypass / resonance
Temperature effect on DCR:
ΔR/ΔT = +0.39%/°C (copper)
At +85°C vs +25°C: DCR increases ~23.4%
→ IL increases at elevated temperature
→ Critical for PoE designs with self-heating
VNA Measurement Setup
Vector Network Analyzer (VNA) S21 measurement:
VNA Port 1 ──[100Ω]── Transformer Primary ──┐
│ (magnetic coupling)
VNA Port 2 ──[100Ω]── Transformer Secondary ─┘
S21 (dB) = forward voltage transmission = Insertion Loss
Sweep range: 100 kHz – 200 MHz
Termination: 100Ω differential at both ports (simulates cable/PHY impedance)
Level: 0.1V RMS or per manufacturer spec (keep below saturation)
Without 100Ω termination: results are meaningless
→ Floating secondary = very low insertion loss (no load, no loss path)
→ Mismatch at primary = resonances not visible in real circuit
1000BASE-T: Pair Balance Requirement
All 4 pairs must be measured individually:
Pair 1: IL = 0.45 dB
Pair 2: IL = 0.48 dB
Pair 3: IL = 0.51 dB
Pair 4: IL = 0.44 dB
Max variation: 0.51 - 0.44 = 0.07 dB ✅ (well within 0.5 dB limit)
Why it matters:
PAM-5 receiver sees different signal amplitude on each pair
→ PHY DSP compensates, consuming noise margin
→ Large imbalance → reduced link margin → packet errors at max cable length
Datasheet Evaluation Checklist
✅ Insertion loss vs. frequency curve (not just single-frequency value)
✅ Maximum guaranteed value (not just "typical")
✅ Test conditions stated: frequency range, termination impedance
✅ Temperature range (25°C? full operating range?)
✅ Pair-to-pair balance data (1000BASE-T)
❌ Red flag: "IL < 1.0 dB typical" with no curve, no conditions
❌ Red flag: Specified at one frequency only (e.g., "< 1.0 dB @ 10MHz")
❌ Red flag: No maximum value — only typical
Source
Voohu Technology (www.voohuele.com) — LAN magnetics and network transformers with full insertion loss characterization. MOQ 50pcs, DHL 3–5 days.
Top comments (0)